Diagnostic extended usefulness of RMI: comparison of four risk of malignancy index in preoperative differentiation of borderline ovarian tumors and benign ovarian tumors

The concept and treatment of borderline ovarian tumors are in controversial for more than a century. Borderline ovarian tumors (BOTs) could form a separate entity that different with benign and malignant ovarian neoplasms. These tumors are histopathologically different by abnormal epithelium and may become cancer. Hence it is also called “ovarian low malignant potential tumor”, as those tumors are believed to have characteristics related to invasive ovarian cancer [1]. It was first described by Taylor in 1929 and officially classified by the International Federation of Gynecology and Obstetrics (FIGO) in 1971 and World Health Organization (WHO) in 1973 [2‐4]. These tumors account for approximately 10–20% of all ovarian epithelial tumors, especially in women of reproductive age [1, 5]. So far, six subtypes of BOTs are identified as: serous (50–55%), mucinous (30–45%), endometrioid, clear cell, seromucinous and borderline Brenner tumor of the ovary [6].

Current findings suggested that the serous borderline ovarian tumors (SBOTs) have more potential to develop into low-grade serous carcinoma, while other borderline ovarian tumors present relative “inert” behavior [7]. Based on this conception, grouping BOTs into different histological subtype and distinction from benign ovarian tumors is of great translational research interests. The distinction of borderline from benign is important since the recommended surgery method is completely different, besides conservative fertility treatment [8]. As lacking effective indicators for preoperative diagnosis and with economic considerations, clinicians would not decide to send samples for an intraoperative frozen section examination if the tumor looks like “Benign” before the operation, which could make the clinical situation into a dilemma for a secondary surgery.

As BOTs have less distinct ultrasound characteristics, other preoperative examinations such as magnetic resonance imaging (MRI), computed tomography (CT), serum levels of CA125, CA199, and even biopsy are often not easy for a definitive diagnosis respectively [9‐15]. However, precise preoperative evaluation of ovarian masses is important to decrease unnecessary anxiety and enable decisions for optimal treatment, especially for patients who wish to preserve their reproductive capacity and do not wish to take a secondary surgery. Thus, specific and sensitive methods for preoperation diagnosing ovarian borderline tumors are needed.

So far, there are only a couple of reports about evaluating the effectiveness of methods in the distinction between BOTs and benign ovarian tumors [16‐18]. The risk of malignancy index (RMI) is probably the most commonly accepted and easy model [19]. RMI is an algorithm based on scores derived from ultrasound variables, menopausal status, and serum CA125 level. Till now, four versions, RMI I, II, III, and IV have been established and generally accepted by clinicians to distinguish malignant ovarian tumors from benign ones.

Our study was purposed to evaluate the availability and performance characteristics of the four RMIs to discriminate BOTs from benign ovarian tumors. Also, we are trying to provide an effective preoperational evaluation module between benign and borderline ovarian tumors in histological subgroups in order to facilitate clinicians choosing a best therapeutic strategy for patients.

In the 1990s, Jacobs et al. originally developed the RMI, which is known as RMI I [22]. Modifying RMI, Tingulstad et al. developed RMI II and III, with the alternation of the ratio of ultrasound score and postmenopausal status score [23, 24]. Recently RMI IV was created by Yamamoto et al. by adding the parameter of the tumor size [25]. Over the past few years, the performance of RMI to distinguish benign from malignant adnexal masses has been well studied. However, how to discriminate borderline ovarian tumors from benign ovarian tumors has been in great difficulty over years, as BOTs present less typical tumor features [26, 27]. In fact, the preoperative discrimination is quite important for BOTs, as the recommended surgery methods are different (Fig. 3). Our study has revealed the effectiveness of using RMIs to predict tumor nature, which could help both surgeon and pathologist making pre and in operation decision for proper treatment to benefit patients, especially who wish to preserve their reproductive capacity before the operation.

In previous studies, BOTs are not evaluated as a separate group and usually included in malignant groups, but their clinical features are more easily to be confused with benign ones. Although the clinical outcome is good, there are still many advanced cases. For the reason above, we applied these RMIs only between BOTs and benign lesions to assess RMIs performance in the differential diagnosis. Our results show that RMI I conducted the best performance in BOTs group, SBOT, and MBOT subgroups. The AUCs of the RMI I were 0.825, 0.839 and 0.791 respectively. It suggests that RMI I was the best method to differentiate BOTs from benign ovarian tumors. Moreover, we found that the AUCs of four RMIs in BOTs and SBOT group were both more than 0.7, it implies that RMIs are possible to identify SBOT before the operation. However, in MBOT group, the AUCs of four RMIs were smaller, especially for the RMI II and RMI III, which were both less than 0.7. Gotlieb et al. showed elevated CA125 concentrations in 75% of SBOT and only 30% of MBOT [10]. This may partly account for the poor performance of RMIs in discriminating MBOTs and benign ovarian masses. Regards of the sensitivity, we found RMI II was the highest for BOTs group, SBOT, and MBOT subgroups. However, there is a risk of use RMI II, as it provides more weighting to the ultrasound findings when compared to RMI I, RMI III and RMI IV. This also explains the improved sensitivity in RMI II. In MBOT subgroup, the sensitivity of RMI II and RMI IV were similar and better than other groups. The most significant factor is that RMI IV included a new parameter about the tumor size. From the previous study, we know that MBOTs demonstrate a significantly larger tumor size than SBOTs [28]. Taken all together, the specificity of RMI I was the highest in all the three groups. The cutoff of the previous studies which investigated the difference between benign and malignant ovarian tumors is 200 for RIM I, RMI II and RMI III [22‐24]. The suggestive cutoff for RMI IV is 450 [25]. However, in our study, all the values of the cutoff for the four RMIs are relatively lower. The main reason is that the ultrasound score, CA125, the percent of postmenopausal status and tumor size of BOTs are lower than those of malignant ovarian tumor. The cutoff of RMI I, II and III is about 60, and 100 for RMI IV. As RMI I may take the best performance of distinguishing BOTs from benign tumors, considering its application in malignancy, we may use < 60, 60–200, > 200 as warning lines for clinicians.

Since elevated levels of CA19–9 have been reported in BOT, especially in mucinous histological types [10, 27], measurement of CA19–9 has been proposed to be of some clinical value in combination with CA125 as a marker for serological monitoring of BOT [29]. Accordingly, in some institutions, CA19–9 has been incorporated as a tumor marker for evaluation of patients with adnexal masses. However, none of the national guidelines, including those of the American College of Obstetricians and Gynecologists (ACOG), the Society of Gynecologic Oncologists (SGO) and National Institute for Health and Clinical Excellence (NICE), have included CA19–9 measurement as an adjunct in the triage of patients with adnexal masses [30, 31]. Alanbay I et al. have conducted a meaningful study. They modified the RMI IV formulation, replacing CA125 with CA19–9. Then they compared RMI IV (CA125), RMI IV (CA19–9), serum CA125 and CA19–9 level, ultrasound score, and menopausal status between BOTs and benign adnexal masses. They found the sensitivity of CA 19–9 (40%) lower than CA 125(54%). RMI IV (CA125) was found to be the best predictive method for differentiation of BOTs from benign adnexal masses. Replacing CA125 with CA19–9 didn’t affect RMI IV sensitivity and specificity for discrimination between BOTs and benign adnexal masses [17]. It indicates that CA125 is more important in discrimination between BOTs and benign adnexal masses, or it is appropriate for RMI than CA19–9. Moreover, the level of CA19–9 was shown to be high in several benign ovarian findings, especially mature cystic teratomas [32], and even in nongynecological conditions such as rheumatoid arthritis [33]. Several studies found increased CA19–9 levels in 37.4–39.6% of mature cystic teratomas cases [34, 35]. It may affect the accuracy of discrimination between BOT and benign ovarian tumors. From what has been discussed above, we selected CA125 instead of CA19–9 as a one of the parameters of RMI.

The evaluation of strategies for the BOTs has not been considered by histologic subtype in previous studies, or even with results that it is impossible to distinguish benign tumor from BOTs. Our study has its own limitations that we only classify BOTs into SBOT and MBOT subgroups and more in-depth clinical studies with the large patient number should be added for validation. Also, the ultrasound findings are greatly influenced by the sonographer. However, we hope that our study would be able to solve certain preoperation question raised in borderline ovarian tumors, especially as a potent reminder for the clinicians. .